Generative music has removed one of production’s historic limits: the time required to compose, record and finish a track. A short instruction can now create songs, variations, instrumentals and entire catalogues within minutes. This abundance appears democratic, but it enters an economy built on the relative scarcity of listening. The royalty pool does not automatically grow with the number of tracks.
Streaming platforms distribute part of subscription and advertising revenue according to plays considered valid. If thousands of synthetic tracks are uploaded and played by automated accounts, they can capture money that would otherwise reach real artists and producers. The question is not merely aesthetic. It concerns identity, transparency and market integrity.
Tidal has taken a clear position: identify fully AI-generated music, label it and make it ineligible for royalties. The tracks may remain available, but listening is separated from participation in the same economic pool as human work. The decision assigns value not only to the sound that results, but also to the presence of a recognisable creative contribution.
The distinction becomes harder as soon as AI plays a partial role. An author may use it to explore a chord, isolate a voice, create a texture, translate lyrics or restore a recording. Calling the result “AI-generated” does not adequately describe that process. Graduated categories and intelligible disclosures are needed: which part came from a model, which materials were authorised and who made the final decisions.
The dispute between GEMA and Suno shows the other side of the issue: training. A German court found that protected works had been processed without the necessary rights and ordered disclosure of revenue and damages to be determined. The issue extends beyond an obvious imitation. A music model absorbs structures, styles and relationships from vast catalogues, making it difficult to separate statistical innovation from the economic use of existing works.
Licensing may turn conflict into a market. Artists and publishers can authorise the use of compositions, recordings, names or voices under defined conditions, receive compensation and retain the right to refuse. Yet a broad collective licence can obscure the difference between voluntary participation and inclusion by contractual inertia. Opt-in must be genuine, specific and reversible.
Discovery is another scarce resource. Even without fraudulent royalties, synthetic tracks can occupy recommendation slots, playlists and search results. If making a thousand songs costs almost as little as making one, quantity becomes a strategy for capturing attention. Platforms will need to judge not only whether a stream is authentic, but whether distribution systems reward those able to saturate the catalogue.
For independent creators, AI remains a powerful tool. It can help a filmmaker make a temporary score, a musician explore arrangements or an author work without an expensive studio. The aim should not be to prohibit these possibilities, but to prevent accessibility from being financed by hiding the work on which models learned and devaluing those who create music with time, skill and accountability.
Infinite music forces the industry to redefine what it pays for. Counting plays is insufficient when identity, origin and audiences can all be automated. The future of royalties will depend on connecting each track to a declared process, verifiable rights and people able to authorise the use of their work. When production has no limit, trust becomes the truly scarce resource.